Hmm,

www.intellasys.net

An old idea, new designs in the works. but does 24 cores full speed (can't say, but similar) for the same power as one full speed tile core uses on average). Of course it is pretty lightweight in processing power, but it has potential with some minor modifications that have been suggested on their forums, and the comp.lang.forth Usenet forums. 

Apparently 100 chips with 24 cores or so each will fit nicely on a little board, let alone one 100 cores on a small chip. Pricing, don't know, $20, dollars or cents in enough quantity?

At the moment it is aimed more at the lowest embedded micro-controller and fpga markets, apparently with somebody with video and audio codec stuff happening in the background already using the existing designs (whatever they are).

However, I think I read ambarella using an array of processors for their video codec, which is much less power. But if GPU's can already perform complex processing and CPU functionality, why are we bothering, why not just make the smallest ATI/Nvidia chip into a micro-controller (of course it is not the most energy efficient).

neat, but you can buy neither the chip nor the processor board, what's point?

Once again Charles D. has written intresting & fact filed article. So Moving Charles up to Msr. Charles, in line with Msr. Mike &Msr. Hale.

How'd I explain it. Simple. Say computer was lcd , & you could have ordinary analog switch, No Probs, wild & Crazy resistance variations, Surges & Heat. its osciloscope might be:/----____/\. weell, just something simple up & down, cann't seem to find keys for dia. 
However computer would turn light on in digital with thousands of connection steps. each independent, so there preperation for flucuation of lamp, timing, dimmer, less heat both preparing on & off mode & power routing. So more cores is more preperation & more stable throughout. Each adition of cores & ability to respond to software is one more step in maintaing order& quality. Complexity is its most important part.
drashek

That's some really neat stuff. I can see this being really great for simulations - hell of a lot faster than FPGAs. You could gang processors to create variable width vectors or set up sets of processors as virtual domains... very cool.